Can we talk about vibe coding?

If you want an idea of what is possible with VibeCoding, especially from a blind perspective, check out Stephen's thread below. Oh, and also check out the AppleVis podcast that correlates.

https://applevis.com/forum/assistive-technology/stop-waiting-ai-tell-you-what-see-start-exploring-it-yourself
https://applevis.com/podcasts/applevis-extra112-stephen-lovely-rethinking-visual-accessibility-vision-ai-assistant

Nothing worth of real value has ever been produced with vibe coding, it's a total waste of resources using very powerful technology the wrong way just because C-suites want to be self-sufficient. All the AI junkies are collectively digging their career graves, not because AI is getting any better but because they are letting their skills rot away by relegating themselves to riding on the passenger's seat instead of driving innovation as pilots.

Large language models are interesting from a scientific perspective, but in terms of production they are actually contributing negatively to society. The same systems that are being used to power them at a loss and at scale could instead be employed to do a lot more interesting scientific research. At this pace we are more likely to end up in a reality where AI outsmart us not because it evolved to super intelligence but because we got a lot dumber collectively speaking.

vibe coding is great if you just want to get started with an idea and want a proof of concept. However, the "move fast and break things" philosophy has made software far worse because shipping with bugs is now more tolerated than ever. Many accessibility bugs, for example, are produced by the large language model and it doesn't know how to implicitly fix them without human assistance, in my experience.

The limitations are pretty clear to see, as long as you don't buy into the hype. You have already noticed that as complexity increases, so does the amount of errors. Now, we've built over the last year or so very good scaffolding to keep this from happening. Tools like Claude Code, for example, which is not accessible BTW, have a lot of architecture to guide the large language model on what to do. This works, but it still depends a lot on the human using the tool. You will not get a good software project from zero-shot prompting. I don't think this will ever be possible in the future due to the nature of LLMs.

You have to be as explicit as possible with most of them, which is already hard for most humans to begin with. A large part of programming that doesn't get stated is that you often have to translate what the user actually wants into code, but what the user actually wants is not immediately clear by the language that they're using. That's where the human capacity has to come in.

The only frontier where they seem to actually be improving is the mechanics of writing code, which is admirable but won't replace anybody unless they're doing easy tasks and maintenance. The reason people struggle with coding is that they get buried into the mechanics of the thing instead of attempting to understand it from first principles before any coding is actually done. But if all you can do is to measure yourself by the number of errors that you have in your code, which is natural if you're starting, you'll get very frustraded that way. The brain can't keep track of context right away. Rather, it has to adapt over time. LLMs tend to context drift, which is a hard problem to solve. That's why they'll lose track of whatever they're doing unless they are grounded by tricks, either by explicitly reminding them over and over or external memory, which is not always going to work. So really we're trying to hack around their limitations. I like treating them as fancy simulations to bounce ideas off of.

Hope this helps.

Sometime last year, Atlassian broke the UI for a feature I depend on for my work in BitBucket so it was virtually impossible to use with VoiceOver. Well, I have been struggling with it for about 8 months or maybe more and it still doesn't work.

A few weeks ago, I was listening to the developer of Blind RSS on Double Tap talking about Vibe Coding, and it occurred to me that I now have access to a ChatGPT Business Plan.

So I installed "codex" which is the ChatGpt commandline. Connecting to my account was a bit confusing as VoiceOver struggled a little to tell me what my options are, but I got there.

At this point I should tell you I write Python for a living. However, I have never written Swift (the language you use for iOS or the Mac0, and I have never developed a MacOS UI.

I told Codex what I want - a native MacOs app, connecting with BitBucket Cloud, that could help me load up a pull request and view comments. Over the course of 3 days and maybe about 8 hours in total, I had something working that was genuinely useful. I have been refining it a lot since then and adding a load of features, although I still can't do everything I need yet.

But without touching a line of code I have something working and 100% tailored to my use case.

In this scenario all I needed before hand was xcode (free download for MacOs/iOS dev), nodeJs (free install) and codex (free download/install). I needed a paid subscription I believe. There is now a MacOs app for codex which seems accessible from a first play. However whether it stays wthat way is anyone's guess.

There is absolutely no way I would have developed this on my own. Much as I would like to have the time and energy to learn Swift and figure all this out on my own, if I had gone down that path I would never have anything working and the idea would have just faded from view over time.

I did get a little too ambitious to start with. I think keeping it small and iterating probably works better than trying to describe everything in one go. The first version of the app, for example, just had hardcoded values everywhere. I was particularly impressed with the way it created a diff view so I could see what has changed in a file. I described one way to do it which I think it largely ignored and what I got was actually pretty fantastic. Simple but understandable. The best thing I have used natively on a Mac with VoiceOver even if it is lacking in features right now.

There are things it struggles with. For example, I wanted it to build a tree for the files but it seemed to struggle. Maybe it was my prompts. Sometimes it has failed to do something - e.g. I wanted to use APIs to get some extra details for things and it just couldn't do it - it tried and then rewrote and rewrote again and again but every time it failed.

The other thing - I keep telling it to make sure it works with VoiceOver. It generates an agent.md file which is a markdown with instructions about how it should work so I added it there. But I can point out bugs where VO navigation isn't working and it will usually fix them even if it can take a number of goes.

One of the best things I asked it to do was add an HTTP Log so I could see what was going on. That helped me notice some crazy things - e.g. it was going through all pages of an API so calling it loads of times, and then I wasn't even using that data. So I was able to direct me.

I've started to try to pay more attention to what it is doing. It is sometimes adamant that something that doesn't work is correct but it can do a work-around for the instances when it doesn't work, as opposed to just using the right thing. I need to be careful about relying on it too much without paying any attention at all.

I think there is a definite danger to developing an app without actually understanding anything that is going on. The first time it broke the build and I had to figure out how to find the error to paste into it was unnerving. But I have no idea what the code is doing. I have no idea what the UI looks like or how many hacks have gone in to get things to work. Honestly, I don't care because the app is already incredibly useful and will hopefully continue to be.

It would be a little different if I decided to release this into the wild. For example, I am pretty sure the keys I used for the API are not in the code but I haven't checked. And I can't say for sure it's not doing anything it shouldn't.
Obviously it would be better if Atlassian didn't just break everything all the time and then not fix it. I shouldn't have to do this. But I am really grateful for the option.

In terms of a process, I find it both enjoyable and frustrating. It's amazing to say "can you just add a list of such and such containing this..." and then a minute later there it is. That is pretty mind blowing. On the other hand, you do need some patience when you endlessly iterate over a problem and it continues to be incapable of fixing it. I have wasted a lot of time trying and failing to have it do something that feels trivial.

Also sometimes you wonder if it really knows what it is doing. It does occasionally feel quite trial and error.

Oh before I forget, one essential thing you should be familiar with is git, the version control system. Every time codex manages to do something and it seems to work I tell it to commit the changes. You don't have to know the syntax or do it yourself, just be aware that you should do this a lot. The reason being is that if it makes a total mess of things you can always say "revert all changes!" and get back to where you were.

OK the last comment was a bit long, but I thought I would follow-up with how I am using this in my professional life.

Firstly, I think it is a bad idea to write code you do not understand in a professional basis. Let's face it, over the years you do have abstractions so many developers don't understand absolutely everything anyway. E.g. if you are writing a Mac app, you might drag controls onto a design surface but you don't really know how they are getting rendered or you will probably use a library to access a database. But I think it is a big step from there to understanding nothing about your own work.

we have setup codex as part of code review. I am a little uncomfortable sending all our software up to ChatGPT to look at. This doesn't feel sensible but it's not my call. The results, however are actually pretty amazing. It can really understand a code base, and even look at the ticket to try to understand why a change is being made. It often comes up with incredibly insightful comments. It's not always right but it is always worth considering what it says. Compared to human reviews I get which are almost always pretty useless other than pointing out some cosmetic errors.

I will use codex to vibe code a proof of concept, particularly if I am not given the time to do it properly myself. I have only done this once and my intention is to rewrite or refactor so I can make sense of it all as I don't really understand what it is doing yet. Which is OK for a quick POC but not good long-term.

And you can always use it for little bits. For example, I asked it to add a few things to an existing template when I didn't know the syntax. I had a look to see what it had done afterwards and it just saves time when compared to going to google or normal ChatGPT.

I don't really want to use vibe coding for anything I have to support because it is essential I know what is going on. I don't think I would feel comfortable going all-in with vibe coding professionally, but I think there are related tools that are useful and it certainly does have its uses.

I do have a lot of concerns about vibe coding and AI in general. Much as I like some of the tools it gives me, I think the bigger picture is pretty terrifying and I hate to think of the implications it has on society and the environment. It worries me a lot that things are becoming effortless to the point where no one is going to really value anything any more. I I think you get more value when there is effort and graft, blood, sweat and tears. And less when you just say "I want it now!" and there it is and now it's boring and onto the next thing.

But as a blind man I appreciate all the help I can get.

I am of two minds when it comes to VibeCoding. On one hand, I absolutely agree with @João Santos; that fully relying on AI to write code and build programs of any significance, is just going to lead to disaster. As someone who graduated with a degree in computer science, I have an understanding of what it takes to learn code, to learn syntax, and to understand the differences between variables and operators, 'if statements' and boolean states, arguments and definitions, just to name a few things.

On the other hand, I don't think there's necessarily anything wrong with using AI to doublecheck a line or two of code that you are perhaps having trouble with. Say for example if there is a line, or even a block of code that is giving you trouble, perhaps it is producing an error and you can't quite figure out how to go about correcting it, or other such situations. In these instances, I don't see how utilize an AI is necessarily a bad thing.

you need pre-existing knowledge. You can vibe an app and release it, but what will you do if it breaks? how do you know it's secure? And all these self-contained platforms that people are using to vibecode apps. What happens when that company goes down, or you want to move your hosted app to somewhere else?

Where vibecoding can be useful is in situations where you've got a bug in a large codebase and you can't figure it out. Or when you want to automate a load of boring, repetitive tasks. Or when you want a very simple script to do something for yourself.

Case in point. I was working on a mechanical project yesterday, building a turntable part. I wanted to evaluate the fundamental resonance frequency of an aluminium component, and find an optimal value where the resonance frequency was above a certain point, the moment of inertia was within a certain range, the overall component mass was low, but there was still x amount of material left. Because i can't use FEM tools in a CAD program, using Python was the best way to do it. The script had to digitally model the exact part, run through as many combinations of numbers as were valid (nearly 7000 combinations) and output the results. I could probably have written the python script myself in half an hour or so, once I'd figured out the math, looked up the syntax for any unfamiliar functions etc. Gemini wrote it in about 30 seconds, I checked it for any errors, and had my part optimised in less than 10 minutes.. that's where vibecoding can be useful. But not for writing apps, and not if you don't check its work before you run the code.

You might be interested in what amazing stuff Devin is coding
https://tweesecake.social/@pixelate

...take responsibility for your code...

There is no truer statement. Oh, and feel free to be as ambitious as you like, just do it on your own merits. With regards to Xcode, if you are coding on a Mac, you have limited choices here. At least in my experience. You can do XCode, you could do VIM or Nano from within commandline via Mac Terminal, or you can write your code manually in a simple text editor, compile it an Xcode, and go from there. As for where to start, I would recommend something like Python, as Python is more or less a lot more straightforward than some of the other more robust programming languages out there. Sadly, I do not know if the Python IDE is accessible these days, but you can run Python through command line, and test out your code that way without using Xcode, if Xcode is too overwhelming for you.

HTH.

Edited because I forgot to switch to markdown

I agree with a lot of what was said so far. Unless we get a massive shift in how these AI models are created, there will likely never be a situation where you can type a one shot prompt and get a completely working app with zero issues or vulnerabilities. With that being said, these tools are definitely useful and you shouldn't dismiss them. I don't think it's a good idea to do so called "vibe coding" without understanding anything about the underlying concepts behind programming, the language you're writing in, etc., but it can speed you up for specific tasks.

For example, let's say you wanted to add a feature to an app you're working on. If you already have a decent idea of what you want and how it should be implemented, it might be faster and easier to ask one of these coding agents to try and code it for you, of course monitoring as it goes and reviewing the output. For some people, including myself who's pretty new to CS, I find it easier if some code has already been written that I can just review. Sometimes the hardest thing is getting started, and these tools can help out with that problem.

Even companies like Apple are starting to embrace this technology. I watched a developer session a few days ago where they were showing off LLMs directly inside Xcode to build an iOS app. They weren't prompting it for the entire thing, they were going step-by-step and looking at the app take shape in real time.

I would have thought think the starting language will depend on what you want to do with it. If all you are interested in is building a native app then I would likely look to XCode and Swift. As I said before, I only dipped my toe into Swift but I would imagine it is what most people use for apps and there are plenty of blind people using it.

I mostly like Python. It isn't quite as punctuation heavy as some languages which works better for me with VoiceOver. However, not all of its standard methods are easy to hear as it mashes words together, and you must enable indentation sounds of some sort.

If you are happy writing little command-line tools for the terminal then Python will do pretty well. I've never tried writing a UI with it and honestly I'm a bit suspicious of doing so. It can be used for web APIs or server-side applications if that's your thing. Or if you were writing an Alexa skill or some such it would likely be good.

As for IDEs, PyCharm is excellent and works really nicely with VoiceOver. There are a few things that aren't accessible (like the database table viewer which I never use anyway) and the built-in terminal last time I checked. But all the things I use on a daily basis work as well as anything on the Mac.

Personally, and no doubt this will attract some scorn, but I think coding with a pure text editor is a fool's game. I think there are some usable text editors on the Mac but I can't say I love any of them. BBEdit is maybe the best of the bunch. Smultron isn't bad. I use TextMate for a few things but I've never tried actually writing anything in it. I think it's more of a fault with VoiceOver than any of the actual editors but none quite work the right way for me.

Anyway, just my thoughts. But I will admit I am totally stuck in my ways.

Sorry glossed over half the reply. Firstly, never feel bad about asking questions - the ones you are asking are great questions anyway but even stupid ones are fine. I've asked enough myself. If anyone is down on you for having interest and enthusiasm and wanting to learn, then that's their problem.

What do you mean by sequential? You mean you write the code line by line from top to bottom?

You certainly can write a Python script like that. You will want to break into functions and then other modules and other things at some point. So for a small script then you can just batch a load of lines together, but for anything complicated you will likely to need more structure.

Correct me if I'm not really answering the question as I might have misunderstood.

For context, I’m writing this as someone who has an understanding of software architecture but has not written code in over 20 years. I have been using AI Several hours a day for about two years now.

As of today, my opinion is that those who heavily caveat the capabilities of using a large language model for coding are either not using the right tools or do not want to be honest about the capabilities because their paycheck depends on it.

I use Claude Code in the terminal with Opus 4.6. I primarily develop in python and JavaScript because those are the languages that the model has been trained on most extensively. Front end development is possible but The output is generic and sighted people will hate it. But if you are developing a backend for a project using node, it’s extremely powerful. I regularly complete projects, including API integrations, with no errors or bugs.

The criticism that you will not be able to develop software using a one shot prompt is true. It will always be true. Because the model doesn’t know the specifications in which your brain and vision the output. I use a plug-in for Claude code called get shit done that you can find on GitHub. Using this plug-in, I spend about 60% of my time building and refining project specification documentation for the model to use when developing the project. This can’t be skipped.

If you have not valuate the state of vibecoding in the last eight weeks, then you know nothing about the capabilities of the models are evolving that fast′

The actual code writing has quickly become the cheapest input of software development. Product design and product management is, at least right now, the most valuable skill Set outside of marketing.

I hope this is helpful and I wish you luck. I don’t check these forms often, but if anything in this post is beyond your technical understanding, I’d recommend pasting it into ChatGPT and asking whatever questions you have there. It is surely a better partner in this exploration than I am!

Yes, Python reads the way you would read, or the way a screen reader reads. That is to say left to right, and top to bottom. Of course, it gets tricky when you do things like import libraries for various things, or defined functions for a particular feature or action for your software. The code will still initially read left to right and top to bottom, but for example let's say you have a function you are defining online 283, if you have not properly imported your libraries, and/or define said function ahead of time, you'll break your code.

I hope that makes sense.

As of today, my opinion is that those who heavily caveat the capabilities of using a large language model for coding are either not using the right tools or do not want to be honest about the capabilities because their paycheck depends on it.

As Linus Torvalds once said in a different context, "Talk is cheap, show me the code!". AI junkies are all deceiving each other with these claims about how great the technology is, but as I said earlier, people are yet to build anything worth any actual value with it, and the service is being sold at a loss. Therefore, if you truly believe the above quote, point me at that amazing code you got an LLM to write so that I can roast it myself. And yes, I already know what kind of answer I'll get,, which is either that the code is either a trade secret (the convenient excuse that people give most often), or you'll just show me a turd that you think is really good when it's actually a huge spam of extremely poorly structured software that looks pretty with lots of useless comments and creative identifier names.

I use Claude Code in the terminal with Opus 4.6. I primarily develop in python and JavaScript because those are the languages that the model has been trained on most extensively. Front end development is possible but The output is generic and sighted people will hate it. But if you are developing a backend for a project using node, it’s extremely powerful. I regularly complete projects, including API integrations, with no errors or bugs.

Who's the experienced engineer actually reviewing the code to ensure that it doesn't have bogs? Reviewing code is significantly harder than writing it, because the reviewer is not in the same loop as the developer, meaning that they are a lot less likely to understand the developer's train of thought and are thus less likely to be aware of potential flaws in the reasoning. From my observations, current large language models are indeed good at getting the happy path right, but any conditions that steer just a little bit away from it are problematic.

The criticism that you will not be able to develop software using a one shot prompt is true. It will always be true. Because the model doesn’t know the specifications in which your brain and vision the output. I use a plug-in for Claude code called get shit done that you can find on GitHub. Using this plug-in, I spend about 60% of my time building and refining project specification documentation for the model to use when developing the project. This can’t be skipped.

Even with a proper specification you won't get anywhere. Just ask those models to produce a C compiler right out of any of the ISO standards, or write a DNS server out of its original RFC, and you'll see that the only thing they'll do great is spend your subscription money.

If you have not valuate the state of vibecoding in the last eight weeks, then you know nothing about the capabilities of the models are evolving that fast′

It wouldn't be very hard to prove the value of this technology if it actually had any, and if producing proper code does require special skills, that alone is already an admission of failure. The idea that is being sold is that the technology is revolutionizing software engineering by making development more cognitively accessible and fast, yet the amazing software built by large language models, which should be everywhere at this point, is actually nowhere to be found...

The actual code writing has quickly become the cheapest input of software development. Product design and product management is, at least right now, the most valuable skill Set outside of marketing.

I'm talking about actual value, not hype value, and by that I mean writing software that solves actual problems, not software that sells because it integrates or was written by AI. These are completely distinct metrics, and the people who measure value based on hype are the ones trying to justify their paychecks.

The TLDR version of @João Santos' post above:

Do not VibeCode.
Thank you, that is all. 😇

Among all the real devs here I'd rank myself at the botum of the list, I'm just a student.
But even I have enough experience with it to know that most of the time I'd rather take an hour to code it myself... There's a real meme around, whether to take 20 minutes to code it yourself or waste 3 days debugging AI output. And it's mostly true.
Mr Grieves, I'm happy with your result to fix atlassium mess, but I think there's more luck here than usual, as pointed out web and front end especially is the most known to llm.
As for the language debate, it will never end, but just know that overall as much as I hate it there is no wrong choice, just some better than others. IMO go with a statically typed language, it will force you to learn things you'll have to learn anyway. Don't choose javascript, I started with php (yes). Html / css are not languages. Like, purebasic (on windows) would be a bad choice.
Otherwise I agree 99.9% with what @João Santos said, the ai vibe coding hype especially is dangerous at best, way too much resources at a terrible environmental, and let's face it, human, cost, for terrible suboptimal quality at best. I'm talking about the psychological suffering of those poorly underpaid 3rd country "worker" who had to extract / filter out of the models the worse of humanity on the internet. You get the idea? Unrelated to this but I'm pretty sure that sb at openai, according to theories, was more or less killed as he was a whistle blower there for something. Just look at wth the situation with openclaw (or whatever its current name is, if it's still around) it's easier to laugh but it's genuinely terrifying what careless people (and those who trust those careless people) can produce.
Follow the primeTime and other real dev content creator online.

Not traditionally programming languages, but I mean they are still languages. Markup and stylesheet to be more precise. The latter is more or less a scripting language, similar to something like Applescript, or JavaScript.
Just saying... 😝

Good point.